Skip to content

Tableau LOD Expressions

Level of Detail (LOD) expressions compute aggregations at a different granularity than what is shown in the current view. They create subqueries independent of the view context, enabling cross-granularity analysis without complex workarounds.

Key Facts

  • Three LOD types: FIXED, INCLUDE, EXCLUDE
  • LOD creates subqueries independent of the current view
  • FIXED ignores the view completely; INCLUDE adds granularity; EXCLUDE removes granularity
  • FIXED LOD evaluates at context filter level (step 4 in filter order)
  • INCLUDE/EXCLUDE LOD evaluate at dimension filter level (step 5)
  • LOD inside complex nested calculations causes performance issues (extra JOINs)

Patterns

FIXED LOD

Computes at specified dimension(s), completely ignoring the view:

{FIXED [Customer ID] : SUM([Sales])}
-- Total sales per customer, regardless of view

{FIXED [Region], [Year] : AVG([Revenue])}
-- Average revenue per region per year

{FIXED [User ID] : MIN([Event Date])}
-- First event date per user (for cohort analysis)

Use cases: cohort analysis, customer LTV, finding first/last purchase date, per-entity totals displayed in any view context.

INCLUDE LOD

Computes at a finer granularity than the view (adds dimensions):

{INCLUDE [Order ID] : SUM([Quantity])}
-- Per-order sum while viewing by month

The view might show monthly aggregates, but INCLUDE calculates at the order level first, then aggregates up.

Performance note: INCLUDE adds 2 JOINs when used in a complex calculation.

EXCLUDE LOD

Computes at coarser granularity (removes dimensions from context):

{EXCLUDE [Region] : SUM([Sales])}
-- Total ignoring region, while view shows by region

Use case: show % of grand total while broken down by subcategory.

Classic Pattern: Percent of Total

SUM([Sales]) / {FIXED : SUM([Sales])} * 100

The {FIXED : SUM([Sales])} with no dimensions computes the grand total, unaffected by any view-level dimensions.

LOD in Filter Order

Understanding where LOD sits in the filter order of operations:

  1. Extract Filters
  2. Data Source Filters
  3. Context Filters
  4. FIXED LOD evaluated here (with Dimension Filters)
  5. INCLUDE/EXCLUDE LOD evaluated here (with Measure Filters)
  6. Table Calculation Filters

This means FIXED LOD is affected by context filters but NOT by dimension filters, while INCLUDE/EXCLUDE LOD is affected by both.

Cohort Analysis with LOD

-- Cohort month (first activity)
{FIXED [User ID] : MIN(DATETRUNC('month', [Event Date]))}

-- Period since first activity
DATEDIFF('month',
    {FIXED [User ID] : MIN([Event Date])},
    [Event Date])

Nested LOD

LOD inside another LOD is possible but use carefully due to performance impact. Each nesting level adds JOINs to the generated query.

Gotchas

  • Avoid LOD inside complex nested calculations - each LOD adds JOINs: 
    -- Slow (3+ JOINs):
SUM({FIXED DATEPART('year', [Order Date]) : SUM([Sales])})

-- Better: pre-aggregate in data source or use table calculation
  • FIXED LOD is not affected by dimension filters on the view - this can cause confusion when a filter appears to not work on a FIXED calculation
  • Context filters DO affect FIXED LOD - promote a filter to context filter if you need FIXED LOD to respect it
  • INCLUDE/EXCLUDE are more intuitive but slower than FIXED in most cases
  • When possible, materialize calculated fields in Extract before publishing rather than computing LOD at query time

See Also